Documentation
Investment Simulation
The Simulation tab answers two questions before you act: what would this trade do to my portfolio right now, and where does my plan take me over the years. Inputs and results persist across tab switches.
When your portfolio changes: an earlier result is marked out of date. Future simulation recalculates using current holdings; What-if retains your trade inputs and asks you to Simulate again. Late responses from the old portfolio cannot replace the newer result. A saved comparison from the old holdings snapshot is cleared, and refresh during Goal seek Apply keeps the original Undo inputs.
Assistant and external-agent simulation commands apply the requested lookback and horizon to the calculation immediately. A goal request waits for its solved plan; automatic form recalculation does not replace that request while it is running. Commands can return from Historical to DCA or Goal seek using the same authorization.
Backtest recommended trades
The backtest needs a portfolio: on an empty one the Historical type shows how to add yours or load the demo instead of its controls.
Choose Run historical backtest on the Trades tab’s trade plan (it was called Backtest trade plan until 30 September 2026). Simulation opens with Historical selected and runs automatically. One row holds the controls: Trade every N months (monthly by default) or days, In-kind transfers (Strict or Cash equivalent, explained by its ⓘ) and Run backtest. Scenario criterion weights changes the weights for this backtest only; Use saved weights restores your saved criteria and Reset defaults the company-rating defaults. None of these change saved settings or place real trades. After a change, a line says the result is out of date until you run again.
The replay starts with the portfolio reconstructed from dated account transactions a year ago, or the earliest available transaction if newer, and ends yesterday — or on the last day with market data, if even the freshest quote is more than seven days old (a note says so). Current planner settings, targets, tax rules and budgets are fixed retrospective parameters. The replay uses the same simplified planner with recorded external flows; it does not invent deposits from Invest. Holdings at the start and still-saved favourites after their recorded addition dates form the candidate universe; deleted watchlists are not archived.
Each decision uses only prices and reports available by that historical close. SEC financial facts become usable the day after filing, including the versions available then. Orders execute at the next observed daily close within seven days. Current analyst targets, forward estimates, sector classifications, fund look-through and valuation anchors are excluded; missing criteria drop out and the other weights scale up. Missing classification or insufficient quality/valuation evidence leaves the company unrated.
Assets the replay cannot reconstruct are left out, not fatal. The run completes and a Left out list above the figures names each one with a short reason (the first five, the rest behind N more):
- an instrument never matched to market data (for example one kept only by its ISIN) — no market data for this instrument;
- a manually valued Other asset, including one of type Cash — an APY is not a market price;
- holdings inferred from a later account balance rather than recorded;
- under Strict, an asset moved into or out of your accounts during the year; under Cash equivalent, one whose move has no price at most seven days old, or only a price from before a split;
- a holding whose prices stop (a delisting) or have a gap of more than seven days;
- a sale with no earlier purchase on record.
A left-out asset is removed from all three lines — both simulations and the actual portfolio, whose line drops that asset's value day by day — so they stay comparable. It goes out whole, from every account that holds it. On an account with a cash ledger its purchases, sales and dividends count as withdrawals and deposits, since that cash left or rejoined the replayed money. A favourite or target without usable prices is only dropped from the buy candidates. To bring an asset back, repair its data in Settings → Data, choose Cash equivalent for a transfer, or confirm moves between your own tracked accounts as internal transfers in Taxes.
The results lead with eight figures: the final value of Recommended trades, the Actual portfolio and Keep opening holdings, the Difference from actual, the time-weighted return, the maximum drawdown, net deposits and the number of trades. The chart draws the three lines. Both simulations get the same external flows; buys use available cash and sales do not enlarge the buy amount. There is no leverage: a withdrawal larger than a simulated portfolio empties it and the excess is not applied. Splits change quantities and unit costs; dividends enter cash on the ex-date — on a day whose dividend reading is missing, nothing is credited in either simulation and Notes names the asset. Unknown lot costs leave tax coverage unavailable and use FIFO for disposal. Accounts without a cash ledger use recorded trade funding as deposits or withdrawals. An account whose opening cash is negative starts at zero, and its later deposits first make up the shortfall.
In-kind transfers defaults to Strict. Choose Cash equivalent explicitly to include transferred assets: each recorded transfer during the replay becomes an equal cash deposit or withdrawal for the recommended and passive simulations, valued with a historical price on or before the transfer, at most seven days old, converted using FX available on or before that quote date. Transfers add incoming assets to the candidate universe from their recorded arrival date. The passive simulation leaves incoming cash uninvested. The result's date line says transfers as cash, and Transfer valuations lists each transfer's account, quantity, quote date, price and cash amount. These amounts are not original acquisition costs.
Folded sections hold the rest: Criteria (each criterion's weight and
whether its history was used, partial or unavailable, with the reason), Executed
trades (fill and signal dates, reasons), Trading dates (buys,
sells and cash on each date), Final holdings, and How it
works (the budget per trading date and the assumptions, one line each). This is a
reconstruction with explicit coverage, not an archive of exact past app output or vintage
provider data. No current numeric metric fills a historical gap. Jobs can be cancelled
and expire from memory; reloading the app does not preserve a run. If a run disappears
after a server restart, choose Run backtest to start again. Launching
while one is active queues the latest request; it starts when the current run finishes
or is cancelled. The assistant can launch it with backtest_trades and reads
which assets were left out and why.
Historical inputs are cached on the server. Daily prices, corporate actions, exchange rates and SEC company facts are stored separately from temporary simulation jobs and survive server restarts on the deployment's persistent volume. Running another scenario reuses valid cached inputs; the historical-data progress counts preparation of each symbol, including cache reads, so it does not mean every symbol is being downloaded again. Successful SEC documents stay fresh for a day. A failed refresh retains valid older data and retries after ten minutes; it does not turn missing coverage into zeros. Refreshing older dividend coverage preserves any wider cached date range.
Under the strategy blocks sit the two windows the projection is built from — Averages over last N months (the lookback the returns are read off) and Simulate N months ahead — a Run simulation button, and the Benchmarks dropdown: S&P 500, MSCI World, Nasdaq 100, US bonds (AGG), Gold, US dollar, Bitcoin and Diversification, each ticked on or off independently. It is the same compact multi-select the Portfolio tab's Portfolio value over time chart uses for its own benchmarks — one pill button naming your picks (the first two, then a +N count) that opens a checkbox list, every row carrying that line's colour. Under it sits the Lines dropdown — Projected with DCA and Projected without DCA — choosing which projected lines the chart draws (see below). A plain-language line above the chart states the result before you read it — how much you would put in over the horizon, where the plan ends up with and without it, and how much of the difference is simply your own new money rather than growth.
The what-if trade simulator
Simulate any buy or sell — an existing holding or a brand-new asset — or several at once, and see, before placing them:
- each traded position's new weight in the portfolio,
- the effect on the diversification score, overall and per dimension (regions, sectors, sizes, types, currencies),
- the shift in the portfolio's weighted fundamentals (P/E, growth, FCF yield…).
Each line is one trade with its own Buy/Sell, asset, amount and unit (shares or your display currency). + Add trade under the lines adds another (up to 25), the ✕ at a line’s end removes it, and one Simulate runs every line together: sells are applied first, so the cash they raise funds the buys, and the result is one combined picture — a sentence per trade, how the buys are paid for, every traded position’s weight, and the portfolio-wide scores. A trade that cannot be priced, or a sell of something you do not hold, is left out of a several-line run and named under the result rather than stopping the rest.
The trade plan feeds it directly: the plan card’s one Simulate button lands here with the whole plan filled in — one line per buy and sell row, each with its amount — and runs it. Selling a held asset also asks your open tax lots for the time-test hints — how much of the sell is tax-free today, and what matures within a year; on the free plan those lots are behind Pro, and the result says so in one muted line rather than silently showing no hint.
Buys spend your broker cash first, so the simulation matches what would actually happen in your account. That is why Total value stays the same after a buy your cash covers: the cash becomes shares, and nothing is added. It grows only by the new money a buy needs beyond your current cash; the row’s ⓘ says so.
The Simulator
The card’s sliding selector offers Projection, DCA, Goal seek and Historical. DCA and Goal seek are mutually exclusive; Projection runs without purchases. Switching to Historical and back keeps your future plan and inputs. On a phone the four choices stay on one line — they tighten to fit, and on a screen too narrow even for that the strip scrolls sideways rather than wrapping.
Projects your portfolio and the selected benchmarks forward at their lookback-average returns — the lookback window is editable, so you can base the projection on the last year or the whole history. The selected mode and optional withdrawals shape the projection:
With an empty portfolio
You do not need a portfolio to try a plan. Since 16 November 2026 DCA and Goal seek run on an empty portfolio: type a ticker, an amount and a period (or a target), press Run simulation, and the chart projects the plan from zero — the plan’s line, the same contributions put into each benchmark, and the uncertainty band. The line under the controls then reads “Run 16:44 from an empty portfolio”, and the summary says that nothing is invested yet, so only the plan’s assets grow. Each asset is measured over the whole lookback window, also when your portfolio is younger than that window.
Two types need something to work from. Projection carries what you hold forward, and Historical replays your trade plan over your own last year. On an empty portfolio each shows a panel that says so, with Add my portfolio (the drop area and the other ways to add data, opened in place) and Explore the demo; Projection also offers Simulate a DCA plan instead. Periodic selling is switched off, because there is nothing to sell from. Once data is in, the card is the ordinary Simulator, and a plan you had already typed is kept.
Dollar-cost averaging
Add any number of assets, each with its own amount and period ("200 into an S&P 500 ETF monthly, 50 into BTC weekly"). Per-asset average daily/monthly returns are shown so you can sanity-check the assumptions. The plan is seeded with your top-3 holdings and runs automatically the first time.
Periodic selling
Schedule regular withdrawals to model a drawdown phase — the benchmarks mirror the same withdrawals, so the comparison stays fair. A projection reads up to 25 buy rows and 10 withdrawal rows; rows that repeat the same asset and period are combined.
Goal seek
Set a target amount and a horizon, and the simulator solves for how the DCA plan must scale — or what fixed 30-day contribution — reaches the goal. The goal is drawn on the chart, so you can see exactly where the plan crosses it.
Saving, comparing and undoing a run
Since 4 September 2026 a run is not disposable. Save this scenario keeps the run you are looking at — named after its own plan — as a dashed line on the chart and in the legend; every later run is then compared with it in one sentence under the chart (“Saved scenario 200/month into VWCE ends at €412k vs €438k now (+€26k)”), until you press Forget. One saved scenario at a time, so the chart never becomes a thicket.
Apply to plan — which overwrites the DCA amounts with what goal seek solved for — keeps the previous asset rows, typed amounts and periods for Undo, including an originally blank amount. Undo becomes available when recalculation finishes or fails and remains available for six seconds. It restores those rows and re-runs; the target and horizon stay as entered. And while a change is being recalculated, a slim progress strip sits over the existing chart instead of the chart silently being replaced.
The uncertainty band
A projection is not one line. Each projected portfolio line (Projected with DCA, Projected without DCA) is deterministic — today's value compounded at the window's average return. Around it the app draws 400 Monte-Carlo paths at the lookback window's own daily volatility — lognormal daily steps, that same average as drift, and your DCA buys and withdrawals happening on every path. The paths' median (P50) sits on the line rather than defining it: the drift is a geometric average with no −σ²/2 correction, so the two coincide by construction. The shaded band behind the line covers 80 % of those paths (P10 to P90); the denser core is the middle half (P25 to P75). The summary sentence adds “— one likely range (P10–P90) is 0.9M–11M with the plan”, hovering a projected day lists P10 and P90 under the line's value and, on a second row, the P25–P75 core, and the ⓘ spells out the recipe (the volatility annualised) and the one assumption to doubt: that the window's average and its volatility persist for the whole horizon. The Uncertainty band checkbox beside the Benchmarks row turns it off for a cleaner chart — on by default, remembered in your browser; the lines do not change. Benchmarks get no band, and a window with fewer than 60 daily returns gets none and says why. With nothing invested yet there is no portfolio volatility to measure, so the band takes the plan’s own: the daily swings of the assets it buys, each weighted by how much it buys, over the days of the same window on which all of them traded (the ⓘ says which one it used; fewer than 60 such days and there is no band). Cash you hold is carried along unchanged.
Which projected lines to draw
The Lines dropdown under the Benchmarks row — Projected with DCA and Projected without DCA, every row carrying its line's colour — chooses which projected portfolio lines the chart draws. Every run picks the default from the plan it sent: with DCA selected (or a goal-seek plan solved) only Projected with DCA is on; with no buy plan only the plain projection is on, and the “with” row is greyed in the menu because that run drew no such line. After a run you toggle freely — both on to compare, both off to see just the actual line and the benchmarks (the button then reads none). A hidden line takes its uncertainty band with it, the legend lists only the visible lines (and the band row only while a band shows), and the y-axis re-fits what is left; the summary sentence keeps quoting both end values, because it describes the simulation rather than the chart. The choice is not remembered — the next run (a knob change, a portfolio refresh, the Run button) re-derives it. The assistant can set the pair: set_simulation {lines: {withDca, withoutDca}}.
Reading the chart's vertical axis
The axis fits what is on the plot: it runs from zero to the smallest round number that still covers every visible line, the top of a shown uncertainty band and the goal line — never more than one tick step above the data. A fan reaching €477k tops out at €500k rather than €600k, and switching the band off brings the axis down with it, so the portfolio's own history keeps its share of the plot instead of being squashed along the bottom.
A Log scale checkbox sits under the plot's bottom-right corner — the same control the Portfolio tab's Portfolio value over time chart carries. On a logarithmic axis every step is an equal percentage rather than an equal amount, so a compounding projection straightens into a line and the near years read as clearly as the far ones. A log axis has no place for zero, so it starts at the smallest positive value in view. It is off by default, remembered in your browser, and shared by Projection, DCA and Goal seek, which draw the same chart.
The simulation chart also draws the diversification score — historical and projected — so you can see whether a plan concentrates or diversifies the portfolio over time.
Projections extrapolate past average returns; they are scenarios to reason about, not forecasts. The assistant can discuss the simulation you just ran — it's part of its context (see The Assistant).
What-if refreshes its sell-lot access state for a changed portfolio, discards obsolete permission errors and shows no directional arrow when a score is unchanged.